EUROPEAN
CENTRAL
BANK
WO R K I N G PA P E R S E R I E S
WORKING PAPER NO. 165
THE INDUSTRY EFFECTS OF
MONETARY POLICY IN THE
EURO AREA
BY GERT PEERSMAN
AND FRANK SMETS
August 2002
EUROPEAN
CENTRAL
BANK
WO R K I N G PA P E R S E R I E S
WORKING PAPER NO. 165
THE INDUSTRY EFFECTS OF
MONETARY POLICY IN THE
EURO AREA
BY GERT PEERSMAN
AND FRANK SMETS*
August 2002
* Gert Peersman: Bank of England and Ghent University (FWO post-doc fellow), email: [email protected]. Frank
Smets: European Central Bank, CEPR and Ghent University, email: [email protected]. Gert Peersman worked on this paper while
being in the ECB’s Graduate Research Programme. We thank Annick Bruggeman, Paul De Grauwe, Geert Dhaene, Freddy Heylen,
Gabriel Perez-Quiros, Rudi Vander Vennet, Anders Vredin, Jan Smets and seminar participants at Ghent University, the Sveriges Riksbank,
the European Central Bank, the European Commission, the New York Fed and the EEA Annual Congress in Venice for many useful
comments.
© European Central Bank, 2002
Address
Postal address
Telephone
Internet
Fax
Telex
Kaiserstrasse 29
D-60311 Frankfurt am Main
Germany
Postfach 16 03 19
D-60066 Frankfurt am Main
Germany
+49 69 1344 0
http://www.ecb.int
+49 69 1344 6000
411 144 ecb d
All rights reserved.
Reproduction for educational and non-commercial purposes is permitted provided that the source is acknowledged.
The views expressed in this paper are those of the authors and do not necessarily reflect those of the European Central Bank.
ISSN 1561-0810
Contents
Abstract
4
Non-technical summary
5
1.
Introduction
6
2.
The industry effects of monetary policy
2.1. Methodology
2.2. Estimation results
9
9
12
3.
Industry characteristics and the monetary policy effects
3.1. Industry characteristics
3.1.1. The conventional interest rate channel
3.1.2. The financial accelerator channel
3.2. Specification and results
15
15
15
17
20
4.
Robustness of the results
4.1. One-step estimation
4.2. Some modifications to the basic model
25
25
26
5. Conclusions
29
Appendix
30
References
32
European Central Bank working paper series
34
ECB • Working Paper No 165 • August 2002
3
Abstract
We first estimate the effects of an euro area-wide monetary
policy change on output growth in eleven industries of seven
euro area countries over the period 1980-1998. On average the
negative effect of an interest rate tightening on output is
significantly greater in recessions than in booms. There is,
however, considerable cross-industry heterogeneity in both the
overall policy effects and the degree of asymmetry across the two
business cycle phases. We then explore which industry
characteristics can account for this cross-industry heterogeneity.
Differences in the overall policy effects can mainly be explained
by the durability of the goods produced in the sector. In contrast,
differences in the degree of asymmetry of policy effects seem to
be related to differences in financial structure, in particular the
maturity structure of debt, the coverage ratio, financial leverage
and firm size.
Key words: monetary transmission mechanism; euro area;
financial accelerator
JEL-classification: E4-E5
4
ECB • Working Paper No 165 • August 2002
Non-technical summary
In this paper we estimate the effects of an euro area-wide monetary policy change on
output growth in eleven industries of seven euro area countries (Austria, Belgium,
France, Germany, the Netherlands, Italy and Spain) over the period 1980-1998. On
average we find that the negative effect of an interest rate tightening on output is
significantly greater in recessions than in booms. There is, however, considerable
cross-industry heterogeneity in both the overall policy effects and the degree of
asymmetry across the two business cycle phases.
This paper explores which industry characteristics can account for this heterogeneity.
We find evidence that differences in the average policy sensitivity over the business
cycle can mainly be explained by the durability of the goods produced in the sector,
and some indication that the capital intensity of production and the degree of
openness have an influence on this average policy sensitivity. This can be regarded
as evidence for the conventional interest rate/cost of capital channel of monetary
policy transmission. These effects are also economically important. The impact of
monetary policy on industries producing durable goods is almost three times as high
than the impact on non-durable goods. However, these interest rate channel
characteristics can not explain why some industries are more affected in recessions
relative to booms than others.
Cross-industry differences in the degree of asymmetry of policy effects over the
business cycle seem to be mainly related to differences in financial structure and firm
size. In particular, we find that a higher proportion of short-term debt over total
debt, a lower coverage ratio, higher financial leverage and smaller firms are
associated with a greater sensitivity to policy changes in recessions. Also these effects
are economically significant. This finding suggests that financial accelerator
mechanisms can partly explain why some industries are more affected in recessions
than others.
ECB • Working Paper No 165 • August 2002
5
1. Introduction
There is a large literature that compares the macroeconomic effects of a change in
monetary policy in the various euro area countries.1 Much less comparative
empirical work has been done based on sectoral or microeconomic evidence.
Nevertheless, such evidence is important as it may improve our understanding of
why the macroeconomic policy effects are different across countries. For example,
Carlino and DeFina (1998) have argued that differences in the regional effects of
monetary policy in the United States are related to the industry composition of the
various US states. Similarly, it has been argued that differences in financial
structure could lead to asymmetries in the transmission process as some countries
are more affected by financial accelerator mechanisms than others.2 Typically, such
transmission channels imply that monetary policy has distributional effects, which
can most easily be detected using dis-aggregated data.
In this paper we analyse the effects of a common monetary policy shock in eleven
manufacturing industries in seven countries of the euro area (Austria, Belgium,
France, Germany, the Netherlands, Italy and Spain). First, we document the crossindustry heterogeneity of the output effects of an area-wide monetary policy
innovation. Following recent research on cyclical asymmetries in the effects of
monetary policy, we also show that most industries are more strongly affected in
cyclical downturns than in booms. Also in this case, there are, however,
considerable cross-industry differences in the degree of asymmetry across business
cycle phases.
Second, we try to explain the cross-industry heterogeneity on the basis of
individual industry characteristics. Following Dedola and Lippi (2000), it is useful
to distinguish between two broad channels: the interest rate channel and the broad
credit channel. As proxies for the determinants of the interest rate channel, we use
an industry dummy for the durability of the goods produced by the sector,
industry measures of investment intensity and the degree of openness to capture
exchange rate sensitivity. As the traditional interest rate channel is expected to be
operative both in booms and recessions, one should not expect significantly
different explanatory power of these industry characteristics in different stages of
the business cycle.
As proxies for the determinants of the broad credit channel, we construct a number
of indicators that may be associated with the strength of financial accelerator
effects. These indicators include proxies for the size of the firms in the industry and
the financial structure of the industry such as financial leverage, the maturity
structure of debt, the financing need for working capital and the ratio of cash-flow
1
For recent surveys, see Guiso et al (1999) and Kieler and Saarenheimo (1998).
2
See, for example, BIS (1995).
6
ECB • Working Paper No 165 • August 2002
over interest rate payments. In contrast to the traditional interest rate channel,
financial accelerator theories typically predict that monetary policy will have larger
output effects in a recession than during a boom.3 The reason is that the external
finance premium which depends on the net worth of the borrower will be more
sensitive to monetary policy actions during a recession when cash flows are low,
firms are more dependent on external finance and collateral values are depressed.
In sum, we expect the proxies for the traditional interest rate channel to have a
significant influence on the overall impact of policy, but not on the differential
effect across booms and recessions. In contrast, the indicators of financial structure
are likely to explain why some industries are relatively more sensitive to monetary
policy changes in recession versus booms.
This paper is related to at least three strands of the empirical literature on the
monetary transmission mechanism. First, a number of papers such as Ganley and
Salmon (1997), Hayo and Uhlenbrock (2000) and Dedola and Lippi (2000) have
recently examined the industry effects of monetary policy shocks. All these papers
find considerable cross-industry heterogeneity in the impact of monetary policy.
Ganley and Salmon (1997) and Hayo and Uhlenbrock (2000) examine the industry
effects in respectively the United Kingdom and Germany. Our study follows most
closely Dedola and Lippi (2000) who systematically analyse 20 industries in five
OECD countries (Germany, France, Italy, the United Kingdom and the United
States). They find that the cross-industry distribution of policy effects is similar
across countries and that these patterns are systematically related to industry
output durability and investment intensity, and to measures of firms’ borrowing
capacity, size and interest payment burden. In this study we focus on seven
countries of the euro area. In addition, we also analyse explicitly business cycle
asymmetries in the industry effects of monetary policy.
Second, our study is also related to the literature that examines whether monetary
policy has different effects in booms versus recessions (Garcia and Schaller (1995),
Kakes (1998), Dolado and Maria-Dolores (1999) and Peersman and Smets (2001b)).
In a variety of countries, those studies show that monetary policy has stronger
output effects in recessions than in expansions. These studies are, however, not able
to distinguish between various explanations for this asymmetry. In particular, it is
not clear whether the asymmetries are driven by asymmetric financial accelerator
effects or by the fact that the short-run aggregate supply curve is convex as in the
so-called capacity constraint model. In the latter model, as the economy expands,
more firms find it difficult to increase their capacity to produce in the short run. As
a result inflation becomes more sensitive to shifts in aggregate demand at higher
rates of capacity utilisation. Using the cross-industry variation, our study is able to
test whether indicators of financial structure and average size can explain the
degree of asymmetry.
3
See, for example, Bernanke and Gertler (1989), Gertler and Hubbard (1988), Azariadis
and Smith (1998).
ECB • Working Paper No 165 • August 2002
7
Finally, our study also sheds light on the empirical literature that tries to test the
empirical implications of financial accelerator theories more directly. A number of
studies find that investment of small firms, which are assumed to have less access
to alternative forms of finance, is more liquidity constraint during downturns. For
example, Kashyap, Lamont and Stein (1994) find for the US that the inventory
investment of firms without access to public bond markets was significantly
liquidity-constraint during the 1981-82 and 1974-75 recessions, in which tight
money also appears to have played a role. In contrast, such liquidity constraints are
largely absent during periods of looser monetary policy. Gertler and Gilchrist
(1994), who examined movements in sales, inventories, and short-term debt for
small and large manufacturing firms, confirm that the effects of monetary policy
changes on small-firm variables are greater when the sector as a whole is growing
more slowly. Non-linearity is also detected by Oliner and Rudebusch (1996), who
find that cash flow effects on investment are stronger after periods of tight money.
Finally, for the four largest euro area economies, Vermeulen (2002) provides
evidence that weak balance sheets are more important in explaining investment
during downturns than during upturns.
The rest of the paper is structured as follows. In Section 2, we first discuss our
methodology for estimating the industry effects of a euro area-wide monetary
policy change (Section 2.1). This requires a measure of the euro area wide monetary
policy stance. In addition, we also need a business cycle indicator for the euro area
to test whether the policy effects are different in booms versus recessions. For both
variables we rely on earlier work. We, then, present the estimation results and
discuss to what extent the effects of policy vary across countries, sectors and
business cycle phases (Section 2.2). Next, in Section 3 we discuss the industry
characteristics that we use (Section 3.1) and present the results of the regression
analysis (Section 3.2). We perform a number of robustness checks in Section 4. The
main conclusions of our analysis can be found in Section 5.
8
ECB • Working Paper No 165 • August 2002
2. THE INDUSTRY EFFECTS OF MONETARY POLICY
In this Section we estimate and describe the effects of a euro area-wide monetary
policy shock on output in eleven manufacturing industries in seven euro area
countries (Austria, Belgium, France, Germany, Italy, the Netherlands and Spain). A
list of the manufacturing industries considered is provided in the data appendix.
We also examine to what extent these effects are different in booms versus
recessions.
2.1. METHODOLOGY
In order to derive the output effects of monetary policy, we estimate for each
individual industry i of country j the following linear regression equation:
[1]
∆yij,t = (αij,0 p0,t + αij,1 p1,t ) + φij,1∆yij,t −1 + φij,2∆yij,t −2 +
(1 −φij,1 −φij,2 )(βij,0 p0,t −1MPt −1 + βij,1p1,t −1MPt −1) + εij,t
where ∆yij,t is the quarterly growth rate of production in industry i of country j,
MPt is the monetary policy indicator and p 0,t and p1,t are the probabilities of
4
being in respectively a recession or an expansion at time t ( p 0,t + p1,t = 1 ).
This reduced-form output equation is inspired by the Markov-Switching model
estimated in Peersman and Smets (2001b). Peersman and Smets (2001b) show that
this model is able to capture the effects of monetary policy innovations on output in
the seven euro area countries considered in this study. Compared to the VAR
approach used in Ganley and Salmon (1997), Hayo and Uhlenbrock (2000) and
Dedola and Lippi (2000), the biggest advantage of this specification is its simplicity.
The single equation approach makes it easy to extend the model to distinguish
between business cycle phases. The parameters β 0 and β1 give the long-run
effects of monetary policy on the industry’s output in a recession and an expansion
5
respectively.
In contrast to Dedola and Lippi (2000) who use domestic monetary policy impulses,
we want to analyse the effects of a euro area-wide change in monetary policy on the
various industries. We think this is a useful exercise not only because it more
closely resembles the current policy regime with a single euro area-wide monetary
4
We will treat both the monetary policy innovation and the recession probabilities as
exogenous to output growth in the individual industry.
5
The single-equation approach will also allow us to do the analysis of the cross-industry
heterogeneity of the policy effects in one step using a panel data approach. See Section 4
below.
ECB • Working Paper No 165 • August 2002
9
policy, but also because during most of the sample period domestic monetary
policies in the seven countries considered were to a large extent coordinated
through the participation in the ERM and other fixed exchange rate mechanisms.6
In order to avoid the simultaneity bias which may result from the fact that shortterm interest rates depend on economic activity through the central banks’ reaction
function, we follow Peersman and Smets (2001b) and use the contribution of
monetary policy shocks to the euro area interest rate in an identified VAR as our
measure of monetary policy impulses.7 The identified VAR we use is described in
Peersman and Smets (2001a). Graph 1 plots the historical contribution of the
monetary policy shocks together with the short-term interest rate. From the graph it
is clear that the years 1982, 1987, 1990 and 1992-93 are identified as periods of
relatively tight monetary policy, whereas in 1984 and 1991 policy is estimated to be
relatively loose.
Graph 1
Contribution of the monetary policy shock to the short-term interest rate
1.5
16
14
1.0
12
0.5
10
0.0
8
-0.5
6
-1.0
4
-1.5
2
1980
1983
1986
1989
1992
1995
1998
Note: The shaded area is the contribution of the monetary policy shocks to the short-term
interest rate (left axis); the solid line is the short-term interest rate itself (right axis).
6
This is definitely the case for Germany, France, Austria, Belgium and the Netherlands. It
is less clear-cut for Italy and Spain who went through various periods of floating
exchange rate regimes during the sample. However, even in this case a large component
of monetary policy innovations is likely to be common with the other countries.
7
We use the contribution of the shocks to the interest rate rather than the shocks
themselves because this allows us to cut down on the number of lags.
10
ECB • Working Paper No 165 • August 2002
In order to distinguish booms from recessions, we again follow an area-wide
approach and use the filtered recession probabilities derived in Peersman and
Smets (2001b). Peersman and Smets (2001b) estimate a MSM model jointly for each
of the seven countries in our analysis and show that those seven countries share the
same business cycle. Graph 2 plots the smoothed probabilities ( p0,t and p1,t ),
together with the de-trended industrial output level in each of the seven countries.
The shaded area is the smoothed probability of being in a recession. The main
recessionary periods are from 1980 till 1982 and from 1990 till 1993. Somewhat
more surprisingly also in 1986 and in the second half of 1995 the probability of
being in a recession is relatively high.8
Graph 2
De-trended industrial production and the probability of being in a recession
Germany
Austria
1.00
12.5
1.00
7.5
10.0
5.0
7.5
0.75
0.75
5.0
2.5
2.5
0.50
0.50
0.0
0.0
-2.5
0.25
-5.0
0.00
-10.0
-2.5
0.25
-5.0
-7.5
1980
1982
1984
1986
1988
1990
1992
1994
1996
0.00
1998
-7.5
1980
1982
1984
1986
France
1988
1990
1992
1994
1996
1998
Belgium
1.00
8
1.00
7.5
6
0.75
4
5.0
0.75
2.5
2
0.50
0.50
0.0
0
-2
0.25
-2.5
0.25
-5.0
-4
0.00
-6
1980
1982
1984
1986
1988
1990
1992
1994
1996
0.00
1998
-7.5
1980
1982
1984
1986
Italy
1988
1990
1992
1994
1996
1998
Netherlands
1.00
10
1.00
7.5
8
5.0
6
0.75
0.75
4
2.5
2
0.50
0.50
0.0
0
-2
0.25
-4
-2.5
0.25
-5.0
-6
0.00
-8
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
0.00
-7.5
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
Spain
1.00
7.5
5.0
0.75
2.5
0.0
0.50
-2.5
-5.0
0.25
-7.5
0.00
-10.0
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
Note: Right axis: de-trended industrial production. The shaded areas denote the probability
of being in a recession (left axis).
8
In Section 4.2. we examine the robustness of our results to an alternative business cycle indicator, which is
based on whether the output gap (estimated using a linear trend) is negative or positive.
ECB • Working Paper No 165 • August 2002
11
2.2. ESTIMATION RESULTS
We individually estimate equation [1] for 74 manufacturing industries in the euro
area over the period 1980-1998. The quarterly growth rates of industry output are
taken from the OECD database “Indicators of Industrial Activity”.9
Graph 3
Cross-industry heterogeneity in monetary policy effects
Beta-estimates
16
9
Policy effect in recession
12
8
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
-0.68
-0.65
1.04
-2.95
0.73
-0.50
3.93
Jarque-Bera
Probability
5.72
0.06
Policy effect in boom
8
7
6
5
4
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
-0.20
-0.21
1.70
-3.01
0.96
-0.46
2.81
Jarque-Bera
Probability
2.71
0.26
3
4
2
1
0
0
-3
-2
-1
0
1
-3
10
-2
-1
0
1
16
Differential policy effect
Average policy effect
14
4
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
-0.48
-0.66
2.50
-3.19
1.17
0.47
2.81
2
Jarque-Bera
Probability
2.83
0.24
8
6
12
10
8
6
4
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
-0.47
-0.49
0.86
-2.42
0.61
-0.51
3.61
Jarque-Bera
Probability
4.40
0.11
2
0
0
-3
-2
-1
0
1
2
-2
-1
0
1
Graph 3 plots the distribution of the β-estimates in a boom and a recession, their
difference and a weighted average where the weights are based on the
10
unconditional probability of being in a recession versus in a boom. The weighted
average is a proxy for the overall policy effect. In a recession, 60 out of 74 industries
are negatively affected by a policy tightening, whereas in an expansion only 41
industries are negatively affected. While the average difference between the effect
in a recession versus a boom is clearly positive at 0.48, there are 20 industries in
which the policy effect in a recession is not larger than in an expansion.11 The
9
Estimations are done for 11 industries from 7 countries. 3 industries (all in Belgium) are
excluded because data are only available for a much shorter sample period. See also
appendix 1 for a discussion of the data.
10
The weighted average of the policy effects in booms and recessions is equal to the
estimated policy effects in a regression similar to equation [1] where we do not take into
account different business cycle phases.
11
In case we find a positive effect of monetary policy, however, this effect is never
significant. For the differential policy effect, we have 1 significant positive observation.
12
ECB • Working Paper No 165 • August 2002
correlation between the policy effects in downturns and those in expansions is
surprisingly low at 0.07.
How different are the policy effects across industries and countries? Table 1
provides an estimate of the country and industry effects by regressing the βestimates on a set of country and sector dummies.12 We also report the effects on the
difference and the weighted average discussed above. The parameters on the
country and sector dummies report the deviations from the mean effect. A number
of patterns are clear. First, it appears that both in recessions and in booms the
average policy multiplier is significantly negative. The average effect over the
business cycle is about –0.47. In addition, the degree of asymmetry in booms versus
recessions is very significant. This confirms the results of Peersman and Smets
(2001b) who find a significant degree of asymmetry using country data.
Second, focusing on the country effects, it appears that the overall output effects of
the common monetary policy shock do not seem to differ significantly from the
average effect in the euro area. In contrast, the degree of asymmetry is significantly
higher in Germany and lower in Italy and Belgium. It is important to note that this
is the case even though we control for the industry composition.13 The higher
asymmetry of Germany is consistent with the findings of Peersman and Smets
(2001b). It is interesting to see that controlling for the industry composition, Austria
and the Netherlands are no longer negative outliers in the degree of asymmetry, as
was found in Peersman and Smets (2001b).
Third, looking at the industry effects, it is clear that the overall policy effects are
small in the food, beverages and tobacco (310) and non-metallic mineral products
(360) industries. In contrast, the overall effects are significantly larger in the
fabricated metal products (381), transport equipment (384) and to a lesser extent,
the chemicals (350) sectors. These results are broadly consistent with the findings in
Ganley and Salmon (1997), Hayo and Uhlenbrock (2000) and Dedola and Lippi
(2000). Overall these studies suggest that the durability of the output produced by
the sector is an important determinant of its sensitivity to monetary policy changes.
This is mainly because the demand for durable products, such as investment goods,
is known to be much more affected by a rise in the interest rate through the usual
cost-of-capital channel than the demand for non-durables such as food. For
example, Dedola and Lippi (2000) report that an industry dummy which captures
the degree of durability is highly significant in explaining cross-industry effects. As
will be shown in the next section, also in our data set this durability dummy is
highly significant.
12
We estimated the effect of the country and sector dummies on the policy multipliers in
booms and recessions jointly using SURE methods. Standard errors are White
heteroscedasticity consistent.
13
Note that Belgium is the only country for which three of the eleven sectors are missing.
ECB • Working Paper No 165 • August 2002
13
Table 1 also shows that there is evidence that the degree of asymmetry in the policy
effects differs systematically across sectors. The textile (320) and non-metallic
mineral products (360) appear to be much more sensitive to monetary policy in
recessions than in booms. On the other hand, there is some weak evidence of
cyclical asymmetries in the basic metal (370) and machinery, except electrical (382)
industries.
Table 1
The industry and country effects of monetary policy
Average
Germany
France
Italy
Spain
Austria
Belgium
Netherlands
310
320
330
340
350
360
370
381
382
383
384
2
R
β0
-0.66
(10.4)
-0.41
(2.58)
-0.16
(1.28)
0.24
(1.67)
0.01
(0.03)
-0.14
(1.17)
0.32
(1.46)
0.14
(1.01)
0.59
(4.07)
-0.02
(0.06)
-0.03
(0.16)
0.38
(2.21)
-0.12
(1.00)
-0.23
(1.26)
0.08
(0.29)
-0.41
(2.73)
0.28
(2.09)
0.47
(2.85)
-1.00
(3.36)
0.71
β1
-0.22
(2.46)
0.30
(1.73)
0.24
(1.57)
-0.33
(2.48)
-0.27
(0.82)
0.25
(1.36)
-0.32
(1.09)
0.13
(0.62)
0.72
(3.50)
0.51
(2.22)
0.44
(1.49)
0.13
(0.49)
-0.35
(1.64)
0.75
(4.56)
-0.67
(3.15)
-0.12
(0.64)
-0.44
(1.34)
-0.16
(0.50)
-0.81
(1.75)
0.41
β0-β1
-0.44
(4.11)
-0.71
(4.01)
-0.39
(1.79)
0.57
(2.73)
0.27
(0.73)
-0.39
(1.61)
0.64
(2.00)
0.01
(0.04)
-0.12
(0.75)
-0.52
(2.11)
-0.47
(1.17)
0.25
(1.06)
0.23
(1.05)
-0.98
(3.17)
0.75
(2.61)
-0.29
(0.97)
0.72
(1.75)
0.63
(1.52)
-0.19
(0.36)
0.46
β
-0.47
(8.58)
-0.09
(0.65)
0.01
(0.19)
-0.01
(0.09)
-0.12
(0.70)
0.02
(0.26)
0.05
(0.24)
0.14
(1.25)
0.65
(4.14)
0.21
(1.06)
0.18
(1.36)
0.28
(1.48)
-0.22
(1.70)
0.20
(2.35)
-0.25
(1.25)
-0.29
(3.86)
-0.03
(0.24)
0.19
(1.40)
-0.92
(3.40)
0.66
Note: White heteroscedasticity consistent t-statistics in parenthesis. Country and industry coefficients
are deviations from overall mean.
14
ECB • Working Paper No 165 • August 2002
3. INDUSTRY CHARACTERISTICS AND THE MONETARY
POLICY EFFECTS
In this section, we analyse whether cross-industry differences in the effects of
monetary policy in booms and recessions can be explained by a number of industry
characteristics. Section 3.1. describes the industry characteristics that we will use. In
Section 3.2. we discuss the regression specification and the estimation results. In
Section 4, we will discuss the robustness of these results.
3.1. INDUSTRY CHARACTERISTICS
In this Section we describe the industry characteristics that we will use to try to
explain the cross-industry heterogeneity in policy effects. Since the coefficients
β ij , 0 and β ij ,1 are averages of the industry behaviour over the estimation period, the
industry-specific variables are also measured as averages over the available
period.14
3.1.1. The conventional interest rate channel
As already mentioned, a first variable that we include to proxy the interest
rate/cost of capital channel, is a durability dummy obtained from Dedola and Lippi
(2000), which is 1 if the industry produce durable goods. 15 We expect a stronger
effect of monetary policy on these industries because the demand for durable
goods, such as investment goods, is known to be much more affected by a rise in
the interest rate than the demand for non-durables.
Apart from the durability dummy, we use one characteristic, the industry’s
investment intensity (INV), to describe the strength of the conventional interest
rate/cost of capital channel. This characteristic, measured as the ratio of gross
investment over value added, has also been used by Hayo and Uhlenbrock (2000)
and Dedola and Lippi (2000). It captures the capital intensity of the industry.
Industries that are more capital-intensive are expected to be more sensitive to
changes in the user cost of capital, which itself will depend on changes in interest
rates. Table 2 shows that in our sample the average investment intensity is about
14
This is also done by Dedola and Lippi (2000). The sample period of the estimation is
1980-1998. However, the indicators from BACH are averages over the period 89-96 (the
largest ‘common’ sample for all industries). This methodology means that we implicitly
assume that the ranking of the industries with respect to these variables is constant over
time. A calculation of the rank correlation for the period 1989-1996 gives us values of
0.88, 0.80, and 0.92 for respectively the working capital, the coverage and the leverage
ratio. For some of the firm size variables, we only have data available for all industries
for 1996.
15
For an explanation of the durability dummy, see the data appendix.
ECB • Working Paper No 165 • August 2002
15
14%. There are, however, considerable differences in investment intensity both
across countries and sectors. The investment intensity appears to be particularly
low in Spain. It is also lower than average in the textile industry and, more
surprisingly in the fabricated metal products and machinery sector. In contrast,
investment intensity is relatively high in the basic metal and transport equipment
industries.
Table 2
Industry characteristics: country and industry averages
Average
Germany
France
Italy
Spain
Austria
Belgium
Netherlands
310
320
330
340
350
360
370
381
382
383
384
INV
0.14
(47.98)
-0.02
(4.67)
0.00
(0.33)
0.03
(4.72)
-0.06
(9.40)
0.01
(1.59)
0.03
(2.85)
0.02
(1.85)
0.00
(0.39)
-0.04
(6.22)
-0.02
(1.77)
0.03
(2.55)
0.02
(2.00)
0.03
(5.17)
0.04
(3.70)
-0.03
(3.37)
-0.05
(5.80)
-0.02
(3.42)
0.04
(3.96)
OPEN
2.20
(18.63)
-0.88
(4.18)
-0.89
(4.58)
-0.99
(4.87)
-1.28
(5.58)
-0.03
(0.17)
2.14
(4.35)
1.93
(5.10)
-0.73
(2.47)
1.00
(1.91)
-0.86
(3.42)
-1.01
(2.87)
0.51
(2.63)
-1.15
(4.11)
0.74
(2.51)
-0.96
(3.46)
0.72
(3.28)
-0.25
(1.07)
1.97
(2.79)
FIN
0.72
(119.47)
0.06
(3.66)
-0.06
(4.46)
0.09
(7.96)
0.06
(4.45)
-0.01
(0.44)
-0.05
(4.36)
-0.10
(5.40)
0.02
(1.70)
0.04
(2.81)
-0.06
(2.48)
-0.06
(2.48)
-0.02
(0.93)
-0.07
(6.48)
-0.07
(3.68)
0.07
(5.55)
0.07
(5.55)
0.03
(1.10)
0.04
(1.49)
WOC
0.73
(44.34)
-0.09
(3.17)
0.00
(0.03)
0.33
(9.91)
0.07
(1.40)
0.06
(1.19)
-0.20
(6.40)
-0.17
(3.95)
0.05
(0.59)
0.19
(6.26)
-0.14
(3.09)
-0.14
(3.09)
-0.06
(1.11)
-0.12
(4.13)
0.09
(1.41)
0.04
(1.19)
0.04
(1.19)
0.08
(1.55)
-0.03
(0.38)
COV
3.53
(35.68)
-0.30
(1.96)
0.65
(2.99)
-1.06
(6.92)
-0.99
(5.61)
1.01
(3.19)
-0.79
(3.86)
1.49
(3.96)
0.55
(1.57)
-0.76
(3.70)
0.10
(0.39)
0.10
(0.39)
0.84
(2.62)
1.16
(3.90)
-0.83
(3.64)
-0.29
(1.09)
-0.29
(1.09)
-0.43
(1.22)
-0.14
(0.26)
LEV
0.55
(82.2)
-0.12
(5.69)
0.03
(3.46)
0.09
(6.40)
0.01
(0.32)
0.00
(0.03)
0.02
(1.84)
-0.03
(1.69)
-0.01
(0.74)
0.03
(2.20)
0.02
(0.89)
0.02
(0.89)
-0.09
(5.22)
-0.08
(3.62)
-0.01
(0.33)
0.05
(4.79)
0.05
(4.79)
-0.02
(1.09)
0.03
(0.86)
SIVAS
0.12
(18.99)
-0.06
(5.06)
0.00
(0.35)
-0.05
(3.29)
-0.03
(2.89)
-0.02
(1.40)
0.09
(4.39)
0.07
(3.02)
-0.05
(2.83)
0.09
(4.80)
0.07
(3.84)
0.07
(3.84)
-0.10
(4.35)
0.00
(0.26)
-0.10
(4.81)
0.08
(3.16)
0.08
(3.16)
-0.07
(4.64)
-0.08
(4.60)
SIVAL
0.67
(82.23)
0.11
(6.95)
-0.01
(0.38)
-0.05
(3.12)
0.03
(1.32)
-0.03
(2.05)
-0.10
(3.74)
0.06
(2.07)
0.09
(4.14)
-0.26
(9.93)
-0.12
(3.75)
-0.12
(3.75)
0.20
(6.92)
-0.04
(2.02)
0.24
(8.96)
-0.18
(8.64)
-0.18
(8.64)
0.16
(11.98)
0.21
(5.81)
Note: t-statistics in parenthesis. For an explanation of the variables, see the data appendix.
Country and industry data are deviations from overall mean.
16
ECB • Working Paper No 165 • August 2002
In addition, we also use, as a proxy for the degree of openness of an industry
(OPEN), the ratio of exports and imports over value added. It is not clear what the
expected sign is of the effect of this indicator on the strength of the monetary policy
effect. On the one hand, a more open sector will be less affected by the slowdown in
the domestic economy caused by the tightening of monetary policy. On the other
hand, a policy tightening will generally lead to an exchange rate appreciation,
which reduces the competitiveness of the sector and may have a negative effect on
external demand. One important drawback of the indicator used is that it includes
both euro area and non-euro area trade. As we are analysing the effect of an areawide monetary policy innovation, the ideal indicator should only include non-euro
area trade. However, we have not yet been able to break down industry trade by
country of destination and therefore could not construct such an indicator. As can
be seen from Table 2, the implication of this drawback is that the openness indicator
is on average much larger for the smaller countries (Belgium and the Netherlands)
than for the larger countries. It is nevertheless useful to include this indicator in the
regression analysis, because the country effects will be picked up by the country
dummies that we include in the regression.
As there are no strong a priori reasons why the conventional interest rate channels
would work differently in booms versus recessions, we expect the durability
dummy, investment intensity and openness to have similar effects over the
business cycle.
3.1.2. The financial accelerator channel
The financial accelerator theory of the monetary transmission mechanism states
that asymmetric information between borrowers and lenders gives rise to an
external finance premium, which typically depends on the net worth of the
borrower. A borrower with higher net worth is able to post more collateral and can
thereby reduce its cost of external financing. As emphasised by Bernanke and
Gertler (1989), the dependence of the external finance premium on the net worth of
borrowers creates a “financial accelerator” propagation mechanism. A policy
tightening, will not only increase the cost of capital through the conventional
interest rate channel, it will also lead to a fall in collateral values and cash flow,
which will tend to have a positive effect on the external finance premium.
Moreover, because collateral values and cash flows are typically low in a recession,
the sensitivity of the external finance premium to changes in interest rates will be
higher in recessions. Monetary policy is therefore likely to have stronger effects in
recessions than in booms.16
In order to test whether differences in agency costs can partly explain the observed
cross-industry heterogeneity in policy effects, we use four balance sheet indicators
16
See, for example, Bernanke and Gertler (1989), Gertler and Hubbard (1988), Azariadis
and Smith (1998).
ECB • Working Paper No 165 • August 2002
17
and two indicators capturing the average size of the firms in the industry. The four
financial indicators are a leverage ratio, a coverage ratio, an indicator of the
maturity structure of debt and an indicator of the need for working capital. We
discuss each of them in turn.
Financial leverage (LEV, i.e. total debt over total assets) is a basic indicator of the
balance sheet condition that is commonly used by financial analysts. However, it is
not entirely clear what sign to expect in the analysis below. On the one hand, firms
with high leverage ratios are likely to face greater difficulties obtaining new,
additional funds on the market, especially during recessions. Based on this
argument we expect that there is a positive influence of the leverage ratio on the
differential impact of monetary policy.17 On the other hand, a high leverage ratio
may also be an indication of the indebtedness capacity of firms. For example,
Dedola and Lippi (2000) interpret the leverage ratio as an indicator of borrowing
capacity, consistent with the findings that more leveraged firms tend to get loans at
better terms. In that case, highly-leveraged firms could be less sensitive to monetary
policy changes.
Our second indicator is the coverage ratio (COV, i.e. gross operating profits over
total interest payments), which measures the extent to which cash flow is sufficient
to pay for financial costs and is therefore related to credit worthiness. Firms with a
higher coverage ratio are therefore expected to be less sensitive to monetary policy
changes. However, also in this case high interest payments could be a signal of high
borrowing capacity.
The ratio of short-term over total debt (FIN) attempts to measure the extent to
which a firm has to finance itself short term rather than long term and is therefore
related to its access to long term finance. With imperfect capital markets, we expect
the spending of firms with a higher short-term debt to be more sensitive to interest
rate changes in particular in a recession. Finally, a related indicator is the working
capital ratio (WOC), defined as the ratio of working capital (current assets minus
creditors payable within one year excluding short-term bank loans) over value
added. The working capital ratio captures the extent to which the firm depends on
financing for its current assets. As these assets typically can not be used as
collateral, this variable proxies the short term financial requirement of the industry.
We expect the financial accelerator to be stronger in industries with a higher level
of working capital.
The balance sheet data used to calculate the financial ratios discussed above are
taken from the European Commission BACH-database. This database is
constructed through the aggregation at the industry level of a large number of
17
18
The ratio of financial leverage that we use is total debt divided by total assets. The
coverage ratio is gross operating profits divided by total interest payments. The results
are however robust to alternative definitions of both variables.
ECB • Working Paper No 165 • August 2002
18
individual firm data. An extensive, detailed discussion of the definitions and the
sources of all the variables is in the Appendix. Table 2 gives an idea of the average
value of those indicators and their differences across countries and sectors. It is
worth noting that because accounting data are typically not fully harmonised
across countries, it may be difficult to compare those ratios across countries. In the
analysis below, such systematic differences should be picked up by the country
dummies.
Finally, the size of a firm is often used as an indicator for the degree of asymmetric
information problems in lending relationships. Agency costs are usually assumed
to be smaller for large firms because of the economies of scale in collecting and
processing information about their situation. As a result, large firms can more
easily finance themselves directly on financial markets and are less dependent on
banks. Greater diversification of large firms can also be reflected in a smaller
external finance premium. We thus expect that industries with a higher average
firm size are likely to do relatively better in downturns and be less exposed to the
financial accelerator. In the benchmark model, we use two size indicators. The first
indicator gives the share of firms with a turnover of less than 7 million ECU in total
industry value added (SIVAS). The second indicator focuses more on the
importance of large firms and is given by the share of firms with a turnover in
excess of 40 million ECU in total value-added (SIVAL). Of course, both indicators
are highly correlated. Table 2 shows that on average the share of small firms in total
value added is about 12 percent, while that of large firms is 67 percent. On average,
the share of small firms appears to be relatively larger in Belgium and the
Netherlands than in the other countries. It is quite low in Germany. Regarding the
industry composition, the food sector has the largest share of small firms and the
lowest share of large firms, while the opposite is the case for the basic metal,
electrical machinery and transport equipment industries.
Finally, Table 3 gives the correlation matrix of the various industry characteristics
discussed above. A number of features are worth mentioning. First, there is a
positive correlation between investment intensity and the share of large firms in the
industry. Capital intensive industries also feature a smaller share of short-term debt
in total debt. Second, there does not appear a strong correlation between the size
measures and any of the balance sheet indicators. Finally, as expected, the maturity
structure of debt and the working capital ratio are highly correlated. Also the
leverage ratio and the coverage ratio are highly correlated.
18
This dataset is also used by Vermeulen (2000).
ECB • Working Paper No 165 • August 2002
19
Table 3
Industry characteristics: correlations
INV
INV
1.00
OPEN
-
SIVAS
-
SIVAL
-
FIN
-
LEV
-
COV
-
WOC
-
OPEN
0.33
1.00
-
-
-
-
-
-
SIVAS
-0.18
0.16
1.00
-
-
-
-
-
SIVAL
0.29
0.11
-0.81
1.00
-
-
-
-
FIN
-0.45
-0.29
-0.17
-0.07
1.00
-
-
-
LEV
0.06
-0.03
0.00
-0.25
0.17
1.00
-
-
COV
0.17
0.08
0.08
0.14
-0.27
-0.44
1.00
-
WOC
-0.11
-0.20
-0.30
-0.05
0.47
0.33
0.42
1.00
3.2. SPECIFICATION AND RESULTS
In this Section we analyse more systematically to what extent the industry
characteristics discussed above can explain the cross-country heterogeneity in the
β-coefficients estimated in Section 2.19 To do so, we estimate the following system of
two equations using SURE methods to account for the correlation in the residuals:
[2]
βij,0 = α 0 + αi ,1dumi + α j,2dum j + α k ,3characteristicij,k + ηij ,0
[3]
βij,1 = α 0 + αi ,1dumi + α j,2 dum j + α k ,3characteristicij,k + ηij,1
where dum j and dumi are respectively country and industry-dummies. In all
regressions we include country and industry dummies to take into account
country-specific and industry-specific effects.20 This is important because our
methodology may give rise to spurious industry and country-specific effects. For
example, the monetary policy effects may differ systematically across countries
because our area-wide monetary policy shock is more appropriate for some
19
This two-step methodology is comparable to the one used by Dedola and Lippi (2000). In
a first step, they estimate the total impact of monetary policy on individual industries
using VARs. In the second step, this impact is regressed on typical balance sheet
characteristics of the industries. One difference here is that we estimate the effects on the
policy multipliers in booms and recession jointly.
20
There is, however, one exception. In the equation with the durability dummy, we can not
include industry specific dummies because there would be exact collinearity. We only
include country dummies for these equations.
20
ECB • Working Paper No 165 • August 2002
21
countries than for others. Similarly, industry-specific effects are important to
control for the possibility that the business cycle of that industry is not fully
synchronised with the common cycle.
In addition, we also estimate separately a similar set of equations for the difference
between the policy effects in a boom versus a recession and a weighted average of
those effects. Obviously, this is just a linear combination of the equations [2] and [3]
above. However, it allows us to directly assess which characteristics have a
significant impact on the total effects and which characteristics affect the
asymmetry in the policy effects across business cycle phases.
In Table 4, we report the results of the estimations when we include the durability
dummy, the other interest rate channel characteristics, the balance sheet indicators
and the size variables separately. In each of these regressions, except the ones with
the durability dummies, also the country and sector dummies are included, but not
reported. Several results are worth noting. First, industries producing durables and
industries producing non-durables both react significantly to monetary policy
shocks and have a significant degree of asymmetry. Focusing on the durability
dummy, we find that this dummy is highly significant in explaining the average
impact of monetary policy. Sectors producing durable products are more sensitive
to monetary policy changes. This evidence in favour of the cost-of-capital channel is
consistent with the findings of Hayo and Uhlenbrock (2000) and Dedola and Lippi
(2000). Moreover, this effect is economically significant. The elasticity of industries
producing durable goods is almost three times as high as the elasticity of industries
producing non-durable goods: respectively –0.61 and –0.22. Table 4 also shows that
the durability dummy has no significant impact on the degree of asymmetry. This
finding is in agreement with our conjecture that this determinant of the strength of
the traditional interest rate channel should not have different effects in booms
versus recessions.
Consistent with the findings of Dedola and Lippi (2000), we do not find a
significant impact of the other interest rate channel characteristics. Investment
intensity and openness do not seem to be important in explaining cross-industry
differences in the overall impact of monetary policy. We only find a significant
effect of the degree of openness in recessions. Sectors with a higher degree of
openness appear to be less affected than more closed sectors. This effect is,
however, relatively small. A 10 points percentage increase in openness, measured
as exports and imports over value added, reduces the absolute value of the βcoefficients with only 0.02. To some extent, this small effect may be due to the fact
21
For example, it could be argued that to the extent that the common monetary policy
shock is dominated by the changes in the German interest rate, such a shock could have
been accompanied by a depreciation of the bilateral DM exchange rate of the currencies
of some of the other euro area countries. In that case, one would expect a stronger effect
in Germany than in those other countries.
ECB • Working Paper No 165 • August 2002
21
that our measure of openness also includes trade within the euro area, as discussed
before. The impact of both variables on the degree of asymmetry is, however,
insignificant. We therefore can not reject our hypothesis that the interest rate
channel works similarly whatever the state of the business cycle.
Table 4
Explaining cross-industry heterogeneity in the effects of monetary policy
β0
β1
β0-β1
Interest rate channel: durability dummy (estimation without industry dummies)
Non-durables
-0.45
0.06
-0.51
(4.16)
(0.45)
(4.12)
Durables
-0.79
-0.36
-0.43
(7.51)
(2.52)
(2.57)
Durability dummy
-0.34
-0.43
0.08
(2.29)
(2.14)
(0.38)
Other interest rate channel characteristics
INV
-0.22
(2.02)
-0.61
(7.18)
-0.38
(2.80)
-5.38
(1.43)
0.01
(0.06)
6.02
(1.22)
0.17
(1.44)
-1.97
(0.89)
0.11
(1.26)
3.57
(1.59)
-0.50
(0.59)
-0.23
(1.66)
3.72
(1.91)
-7.70
(2.75)
0.21
(0.25)
0.48
(2.99)
-5.19
(2.09)
-0.85
(0.67)
-0.36
(0.66)
0.05
(0.68)
0.82
(0.71)
-2.45
3.57
-6.02
(1.72)
(2.05)
(2.93)
SIVAL
3.57
-2.35
5.92
(3.47)
(1.47)
(3.16)
SIEM50
0.95
0.07
0.88
(4.98)
(0.15)
(1.73)
SIEM100
0.55
-0.36
0.90
(2.68)
(1.15)
(2.40)
SITU30
0.55
0.02
0.53
(2.26)
(0.05)
(1.17)
Note: White heteroscedasticity consistent t-statistics in parenthesis.
0.22
(0.18)
0.95
(1.03)
0.57
(2.38)
0.15
(0.85)
0.32
(1.64)
OPEN
0.64
(0.22)
0.17
(2.44)
β
Balance sheet indicators
FIN
-4.13
(2.73)
WOC
-0.29
(0.54)
COV
0.26
(2.90)
LEV
-1.47
(1.02)
Various industry size indicators (separate estimations)
SIVAS
Second, in contrast to some of the interest rate channel characteristics, we find no
significant effect of the balance sheet indicators on the total policy effects. However,
consistent with the financial accelerator hypothesis, we do find that weaker balance
sheets imply a significantly stronger policy effect during recessions than during
booms. The financial variables that seem to work most consistently with the
financial accelerator hypothesis are the ratio of short debt over total and the
22
ECB • Working Paper No 165 • August 2002
coverage ratio. While these variables have no explanatory power during booms,
they do explain cross-industry differences during recessions. Moreover, these
effects are economically significant. The difference in ratio between the industry
with the highest short-term debt and the one with the lowest is about 0.14.
According to the estimates reported in Table 4 this could account for a difference in
the estimated policy effects in a recession of about 0.58, which itself has a standard
deviation of about 0.71. Differences in the coverage ratio can explain similar
magnitudes.
A higher leverage ratio also appears to increase the degree of asymmetry between
policy effects in a recession versus a boom. However, in contrast to the other
financial indicators, this is mainly a result of a perverse effect on the policy effects
during a boom (although only at the 10 percent significance level). In particular,
industries with a higher leverage ratio (i.e. higher debt relative to total assets)
appear to be less sensitive to monetary policy innovations during a boom. To some
extent, this perverse effect may be the result of the fact that high leverage maybe an
indicator of good credit standing and high borrowing capacity as mentioned above.
Finally, the bottom panel of Table 4 reports the results of the various size indicators.
Our preferred size indicators (SIVAS and SIVAL) fail to have any significant effect
on the average impact of monetary policy. This is in contrast to the findings of
Dedola and Lippi (2000), who do find a significant effect in their sample on the total
effects. In order to check the robustness of these results, Table 4 also reports
estimations with alternative size indicators. SIEM50 (SIEM100) is a dummy variable
which takes on the value of one when the average employment of the firms in the
sector is greater than 50 (100). These variables are more comparable to the size
variable of Dedola and Lippi (2000), who also used an indicator based on
employment, but less reliable than the others because we had to use two different
data sets to construct this variable for all countries in our sample (see the data
appendix). SITU30 is a dummy variable, which takes on the value of one when the
average turnover of the firms in the sector is greater than 30 million ECU. We do
find a significant impact of SIEM50 on the overall impact, but this evidence does
not appear to be very robust.
The effect of size on the degree of asymmetry is, however, significant in most cases
(only significant at the 10 percent level for SIEM50 and insignificant for SITU30).
This is the result of a highly significant effect in recessions and an insignificant
effect in booms.22 This is a confirmation of the financial accelerator hypothesis.
Industries with firms of a smaller size are more negatively affected by a policy
tightening in recessions versus booms. Again, this is also economically very
significant for all size indicators. For example, the elasticity to a monetary policy
shock in a recession is, for industries with average employment less than 100 or a
22
For SIVAS, however, we also find a significant perverse effect in booms.
ECB • Working Paper No 165 • August 2002
23
turnover less than 30 million ECU, 0.55 higher than other industries, while the
average impact in a recession is –0.68.
Table 5 shows that these results are robust when we include all characteristics in
the same regression equation. Columns (1) to (3) report the results when
respectively SIVAS, SIVAL and SIEM50 are included as a proxy for size. The only
difference is that we find some evidence for a significant influence of the
investment intensity on the differential impact of monetary policy.
Table 5
Explaining cross-industry heterogeneity in policy effects (joint estimation)
(1)
(2)
SIVAL
β0-β1
8.11
(1.97)
0.08
(0.91)
-6.48
(2.37)
-0.25
(0.36)
0.46
(3.28)
-5.31
(2.44)
-4.59
(2.75)
-
β
-2.31
(0.98)
0.12
(1.45)
-0.56
(0.53)
-0.53
(0.99)
0.02
(0.33)
0.84
(0.83)
0.24
(0.22)
-
SIEM50
0.64
INV
OPEN
FIN
WOC
COV
LEV
SIVAS
2
R
(3)
β0-β1
7.54
(1.83)
0.07
(0.81)
-6.83
(2.50)
0.17
(0.24)
0.45
(3.33)
-4.32
(1.90)
-
β
-2.96
(1.20)
0.11
(1.26)
-0.36
(0.34)
-0.44
(0.87)
0.02
(0.26)
1.37
(1.39)
-
β0-β1
8.93
(2.21)
0.08
(0.81)
-7.37
(2.80)
0.18
(0.30)
0.45
(3.22)
-5.35
(2.58)
-
β
-2.93
(1.23)
0.08
(0.88)
-0.65
(0.52)
-0.10
(0.25)
0.00
(0.03)
1.55
(1.39)
-
1.21
(1.28)
-
-
-
-
3.63
(2.38)
-
0.69
0.64
0.70
0.43
(1.18)
0.62
0.62
(2.78)
0.73
Note: Each regression also includes country and sector dummies. White heteroscedasticity
consistent t-statistics in parenthesis.
24
ECB • Working Paper No 165 • August 2002
4. ROBUSTNESS OF THE RESULTS
In this section, we provide a robustness analysis of the results. Four alternatives are
considered. The first is based on a one-step methodology and is discussed in the
next subsection. The three others, discussed in Section 4.2, are alternatives based on
some modifications of the basic model: different monetary policy shocks, the
contribution of area-wide monetary policy shocks to the individual country interest
rates and the asymmetric effects of monetary policy depending on the output gap
instead of output growth.
4.1. ONE-STEP ESTIMATION
In the Sections above we have used a two-step methodology whereby in the first
step, we estimate the policy effects and in the second step we try to explain the
cross-industry differences on the basis of industry characteristics. In this section we
check the robustness of this two-step methodology by performing the estimations
in one step using standard panel data techniques.
Since p1,t −1 = 1 − p 0,t −1 , we can rewrite equation [1] as follows:
[4]
∆yij,t = (αij,0 −αij,1) p0,t + αij,1 + φ1∆yij,t −1 + φ2∆yij,t −2 +
[1−φ1 −φ2 ][(βij,0 − βij,1)p0,t −1MPt −1 + βij,1MPt −1] + εij,t
where we also have assumed that the autoregressive parameters are the same
across industries. We can now substitute equations [2] and [3] directly into equation
[4] and estimate this equation in one step for all industries simultaneously.23 Table 6
reports the results of a Feasible GLS estimator, which allows for heteroscedasticity
and cross-sectional correlation of the residuals. The latter is appropriate as output
growth is likely to be correlated across industries.
Table 6 shows that the results obtained above are generally robust. We still find that
the durability of the goods produced mainly affect cross-industry differences in the
overall policy effects, whereas the balance sheet indicators significantly affect the
differential policy effect in recessions versus booms. There are two slight
differences with the results reported above. First, using the panel data techniques
the leverage ratio has a significant policy effect in a boom. A higher leverage is
associated with a smaller sensitivity to monetary policy shocks in a boom. As
discussed above, this may be due to the fact that firms with a high leverage are also
firms with a good credit standing. This finding is consistent with the finding of
Dedola and Lippi (2000). The negative effect on the degree of asymmetry, is
23
In order to have a balanced panel data set, we excluded Belgium from the analysis. This
leaves us with 66 industries and 79 periods.
ECB • Working Paper No 165 • August 2002
25
consistent with our conjecture that it is difficult for these firms to get additional
loans in a recession. Second, one of our two preferred size variables (SIVAS) has
significantly the wrong sign in a boom. This would indicate that large firms are
more sensitive to monetary policy shocks in a boom. This finding is puzzling and
we do not have an explanation for this.
Table 6
Panel data estimation – Feasible GLS
Durability dummy
INV
OPEN
FIN
WOC
COV
LEV
SIVAS
SIVAL
β0
-0.31
(3.44)
-0.42
(0.20)
0.13
(2.06)
-2.61
(1.96)
-0.27
(0.67)
0.28
(3.20)
-0.56
(0.49)
0.01
(0.01)
2.99
(2.81)
β1
-0.45
(4.02)
-2.93
(1.15)
-0.01
(0.16)
1.85
(1.13)
-0.33
(0.65)
-0.09
(0.87)
4.18
(3.02)
4.00
(3.33)
-1.21
(1.40)
β0-β1
0.14
(0.93)
2.51
(0.76)
0.14
(1.42)
-4.46
(2.09)
0.05
(0.08)
0.37
(2.67)
-4.74
(2.63)
-3.99
(2.55)
4.20
(3.04)
Note: t-statistics in parenthesis
4.2. SOME MODIFICATIONS TO THE BASIC MODEL
In the basic model, the monetary policy shocks are obtained from a VAR using a
standard Choleski decomposition comparable to the one of Christiano, Eichenbaum
and Evans (1998) for the US. Peersman and Smets (2001a) also present the results of
an alternative identification strategy, similar to Sims and Zha (1998), with a
contemporaneous interaction between interest rate and exchange rate. Moreover,
monetary authorities do not react within the period to output and price movements
because of information lags. The results of the estimates, if we use the contribution
of these monetary policy shocks to the interest rate, are presented in the first
columns of table 7.24
The conclusions are very similar to our basic analysis. The durability of the goods
produced is still an important determinant for the total impact of monetary policy,
and balance sheet characteristics of the firms have a significant influence on the
24
26
We only report the results for the degree of asymmetry and the average impact. The
coefficients in recessions and expansions are, however, available on request.
ECB • Working Paper No 165 • August 2002
degree of asymmetry. The significance of some variables is, however, slightly less.
This is the case for the durability dummy on the total impact, and the debt (FIN)
and leverage ratio on the degree of asymmetry. These variables are only significant
at the 10 percent level.25
Table 7
Results with modifications to the basic model
Other monetary policy
shocks
Contribution to
domestic interest rate
Output gap as business
cycle indicator
β0-β1
β0-β1
β
β
Non-durables
-0.46
-0.18
-0.70
-0.32
(4.39)
(1.88)
(3.93)
(2.46)
Durables
-0.73
-0.41
-0.99
-0.62
(4.35)
(5.50)
(4.40)
(5.56)
Durability dummy
-0.27
-0.23
-0.29
-0.30
(1.40)
(1.87)
(1.04)
(1.77)
INV
2.00
-1.98
3.43
-4.44
(0.53)
(1.04)
(0.50)
(1.75)
OPEN
0.13
0.11
0.22
0.21
(1.34)
(1.58)
(1.47)
(2.82)
FIN
-4.45
-1.49
-4.40
-1.13
(1.82)
(1.20)
(1.20)
(0.70)
WOC
0.26
-0.38
0.36
-0.09
(0.42)
(0.63)
(0.37)
(0.18)
0.08
0.59
0.08
COV
0.36
(2.87)
(1.18)
(3.04)
(0.70)
LEV
-3.51
0.74
-4.83
1.34
(1.75)
(0.64)
(1.57)
(0.91)
SIVAS
-3.63
-0.19
-7.31
0.78
(2.23)
(0.19)
(2.40)
(0.45)
SIVAL
4.03
0.95
6.02
0.49
(3.00)
(1.15)
(2.28)
(0.39)
Note: White heteroscedasticity consistent t-statistics in parenthesis
β0-β1
-0.31
(1.20)
-0.25
(0.97)
0.06
(0.18)
-5.79
(0.98)
0.05
(0.24)
-4.07
(0.86)
-2.07
(1.16)
-0.22
(0.92)
-0.66
(0.23)
-7.91
(2.49)
2.59
(1.28)
β
-0.17
(1.48)
-0.56
(6.50)
-0.38
(2.67)
-1.92
(0.78)
0.11
(1.15)
-1.73
(1.26)
-0.58
(0.88)
0.04
(0.49)
0.32
(0.24)
0.25
(0.20)
1.11
(1.20)
As already mentioned, monetary policy effects may differ systematically across
countries because area-wide monetary policy shocks are more appropriate for some
countries than for others. This should be captured by the country-specific dummies
in the basic model. An alternative is using the contribution of area-wide monetary
policy shocks to the individual country interest rates in addition to country
dummies. The estimation of the contribution of area-wide monetary policy shocks
on individual country interest rates is done in Peersman (2000) by using a twoblock structured VAR with an area-wide and country-specific block. The results of
our two-step methodology, with the contribution to the individual country interest
rates, are reported in columns 3 and 4 of table 7.
25
The ratio of short-term over total debt (FIN) is, however, still highly significant in a
recession, but not reported in the table.
ECB • Working Paper No 165 • August 2002
27
The main results are generally robust. There are, however, some slight differences.
The total impact of monetary policy on industries producing durable goods is still
much larger (double), but less significant than in the basic model (p-value = 0.08).
The ratio of short-term debt over total debt is now insignificant in explaining crossindustry differences in the degree of asymmetry. On the other hand, our openness
indicator has a highly significant influence on the total impact of monetary policy.
Industries with a higher degree of openness are less affected than more closed
industries. Moreover, we also find that industries with a higher investment
intensity are more sensitive to monetary policy changes. The investment intensity
indicator is only significant at the 10 percent level, but highly significant when
estimated in combination with some of the size variables (not reported in the table).
This might indicate that our country-specific dummies in the basic model do not
fully capture the systematic deviation of monetary policy in the individual country.
Finally, we investigate the robustness of our results when we use an alternative
business cycle indicator. So far, we have identified recessions using the filtered
probabilities obtained from a Markov-switching model. A recession is characterised
with, on average, a negative growth rate of industrial production. It is not fully
clear from the Bernanke and Gertler (1989) model, whether we also find an
important role for the financial accelerator theory in explaining asymmetries when
we use the level of the output gap as the business cycle indicator. In order to check
this, we replace the probabilities of being in respectively a recession or an
expansion ( p 0 ,t and p1,t in equation [1]) with a dummy that equals 1 when the
level of the output gap of the individual industry is respectively negative or
positive. This output gap is calculated based on a linear trend. The advantage of
this methodology is that we can calculate this business cycle measure at the
individual industry level.
The results are reported in the last two columns of table 7. The effects on the total
impact are, of course, similar as in the basic model. Interestingly, the average
degree of asymmetry still has the correct sign but is not significant anymore for
industries producing both durable and non-durable goods. This might indicate that
monetary policy only has asymmetric effects depending on the growth rate of
output and not the level of the output gap. Moreover, we do not find any
significant impact of the balance sheet characteristics on the degree of asymmetry
anymore. The only exception is firm size. One of our two preferred size measures is
significant. Also the other size measures, not reported, indicate that firm size can
explain asymmetries depending on the level of the output gap. In sum, we find that
balance sheet characteristics (such as firm size, the ratio of short-term over total
debt, coverage and leverage ratio) are important in explaining asymmetries
depending on the growth rate of output, but only size is important in explaining
asymmetries depending on the level of the output gap.
28
ECB • Working Paper No 165 • August 2002
5. CONCLUSIONS
In this paper we have estimated the effects of a euro area-wide monetary policy
change on output growth in eleven industries of seven euro area countries over the
period 1978-1998. We have shown that on average the negative output effects of an
interest rate tightening are significantly greater in recessions than in booms. There
is, however, considerable cross-industry heterogeneity in both the average policy
effects over the business cycle and the differential impact in recessions versus
booms.
This paper explores which industry characteristics can account for this
heterogeneity. We find evidence that differences in the average policy sensitivity
over the business cycle can mainly be explained by the durability of the goods
produced in the sector, and some indication that the capital intensity of production
and the degree of openness have an influence on this average policy sensitivity.
This can be regarded as evidence for the conventional interest rate/cost of capital
channel of monetary policy transmission. These effects are also economically
important. The impact of monetary policy on industries producing durable goods is
almost three times as high than the impact on non-durable goods. However, these
interest rate channel characteristics can not explain why some industries are more
affected in recessions relative to booms than others.
Cross-industry differences in the degree of asymmetry of policy effects over the
business cycle seem to be mainly related to differences in financial structure and
firm size. In particular, we find that a higher proportion of short-term debt over
total debt, a lower coverage ratio, higher financial leverage and smaller firms are
associated with a greater sensitivity to policy changes in recessions. Also these
effects are economically significant. This finding suggests that financial accelerator
mechanisms can partly explain why some industries are more affected in recessions
than others.
These results are generally robust with respect to an alternative methodology and
alternative monetary policy indicators. However, we do not find an important role
anymore for financial structure variables in explaining asymmetric effects of
monetary policy depending on the level of the output gap. There is only some
indication that firm size can explain these asymmetries. Moreover, the average
degree of asymmetry depending on the output gap is not significant anymore.
Overall, our results are in agreement with those of Dedola and Lippi (2000) who
conclude that there is role for both traditional cost-of-capital channels and the
broad credit channel in explaining the sectoral effects of monetary policy.
Moreover, our results suggest that financial accelerator mechanisms work mainly
during recessions. This is consistent with some of the literature reviewed in the
introduction.
ECB • Working Paper No 165 • August 2002
29
APPENDIX
APPENDIX 1. DATA SOURCES AND DEFINITIONS
Industrial data are quarterly for the period 1980-1998 from the OECD database:
“Indicators of Industrial Activity”. The following industries of each country are
included in our analysis:
•
•
•
•
•
•
•
•
•
•
•
Food, beverages and tobacco (310)
Textile, wearing apparel and leather industries (320)
Wood and wood products, including furniture (330)
Paper and paper products; printing; publishing (340)
Chemicals; chemical, petroleum, coal, rubber and plastic products (350)
Non-metallic mineral products (360)
Basic metal (370)
Fabricated metal products, except machinery & equipment (381)
Machinery, except electrical (382)
Electrical machinery, apparatus, appliances & equipment (383)
Transport equipment (384)
Our estimates concern these eleven industries for the countries Germany, France,
Italy, Spain, Austria, Belgium, and the Netherlands, except for the industries 340,
350 and 383 for Belgium because data are only available for a much shorter sample
period.
The first explanatory variable is a durability dummy, which is 1 if the industry
produce durable goods. This variable is also used by Dedola and Lippi (2000) and is
based on the economic destination of production from the national accounts
statistics. According to this criterion, the ‘durable’ output industries are 330, 360,
370, 381, 382, 383, and 384.
The investment intensity (INV) and openness (OPEN) ratios are constructed from
the STAN-OECD database, which records annual data at the industry level. We use
an average for the period 1980-1996. They are:
•
•
INV: gross investment/value added.
OPEN: (export + import)/value added.
Balance sheet data are from the European Commission BACH-database. It contains
aggregated balance sheets and profit and loss account information at the industry
level. Most of the industries are matching with the OECD dataset, though, there are
some exceptions: Industries 330 and 340 are aggregated in the BACH dataset, as
well as industries 381 and 382. For these industries, the values from BACH are
30
ECB • Working Paper No 165 • August 2002
assigned to both industries. Balance sheet data are averages over the period 89-96
(the largest ‘common’ sample for al industries). The following definitions are used:
•
•
•
•
•
•
Working capital (WOC): the ratio of working capital to value added.
Working capital is defined as the asset item “current assets” minus the
liability item “creditors payable within one year” (except short-term bank
loans). In BACH, this is: (D – F + F2) / T. Results are similar when we
exclude cash and current investment from the ratio, or when we include the
short-term bank loans in the ratio.
Leverage ratio (LEV): ratio of total debt (short and long run) to total assets: F
+ I. Similar results are obtained with the ratio of total debt to capital and
reserves.
Coverage ratio (COV): ratio of gross operating profits to total interest
payments : U / 13. The results are robust to other specifications of this ratio.
Examples are net operating profits or total profits (except depreciations) in
the nominator or total debt in the denominator.
SIVAS (SIVAL): The share of small (large) firms in total industry value
added. These are firms with a turnover of less than 7 million ECU (more
than 40 million ECU).
SITU20 (SITU30, SITU40): is a dummy variable which takes on the value of
1 when the average turnover of the firms in the sector is greater than 20
(30,40) million ECU.
SIEM50 (SIEM100): average employment per firm of the industry. For this
ratio, data is only available for the year 1996 for the industries of Germany,
France, Belgium, and Italy. These data are completed with data form OECD
“Industrial Structure Statistics” for Austria, Spain, and the Netherlands. For
the size variable, we constructed a dummy that takes the value 1 for
industries with an average size larger than 50 (100).
ECB • Working Paper No 165 • August 2002
31
REFERENCES
Artis M, Krolzig H-M. and J. Toro (1999), “The European business cycle”, CEPR
Discussion Paper 2242.
Azariadis C. and B. Smith (1998), Financial intermediation and regime switching in
business cycles”, American Economic Review, 88, p 516-536.
Bank for International Settlements (1995), “Financial structure and the monetary
policy transmission”, C.B. 394, Basle, March.
Bernanke B. and M. Gertler (1989), “Agency costs, net worth and business cycle
fluctuations”, American Economic Review, 79:1, p 14-31.
Calomiris C, Himmelberg C. and P. Wachtel (1995), “Commercial paper, corporate
finance, and the business cycle: a microeconomic perspective”, Carnegie-Rochester
Conference Series on Public Policy, 42, p 203-250.
Carlino G. and R. DeFina (1998), “The differential regional effects of monetary
policy”, The Review of Economics and Statistics, 80(4), p 572-87.
Dedola L. and F. Lippi (2000), “The monetary transmission mechanism: Evidence
from the industry data of five OECD countries”, CEPR Discussion Paper 2508.
Dolado J. and R. Maria-Dolores (1999), “An empirical study of the cyclical effects
of monetary policy in Spain (1977-1997)”, CEPR Discussion Paper 2193.
Ganley J. and C. Salmon (1997), “The industrial impact of monetary policy shocks:
some stylised facts”, Bank of England Working Paper Series, 68, 1997.
Garcia R. and H. Schaller (1995), “Are the effects of monetary policy asymmetric?”,
CIRANO Scientific Series, 95s-6.
Gertler M. and S. Gilchrist (1994), “Monetary policy, business cycles, and the
behavior of small manufacturing firms”, Quarterly Journal of Economics, 59:2, p 309340.
Gertler M. and G. Hubbard (1988), “Financial factors in business fluctuations” In
Financial Market Volatility, Federal Reserve Bank of Kansas City, MO.
Guiso L, Kashyap A, Panetta F. and D. Terlizzese (1999), “Will a common
European monetary policy have asymmetric effects?, Economic Perspectives, Federal
Reserve Bank of Chicago, p 56-75.
32
ECB • Working Paper No 165 • August 2002
Hamilton J. (1989), “A new approach to the economic analysis of nonstationary
time series and the business cycle”, Econometrica, 57 (March), p 357-84.
Hamilton J. and G. Perez-Quiros (1996), “What do the leading indicators lead?”,
Journal of Business, 69:1, p 27-49.
Hayo B. and B. Uhlenbrock (2000), “Industry effects of monetary policy in
Germany”, J. Von Hagen and C. Waller (eds.), Regional aspects of monetary policy in
Europe, Boston, Kluwer, p 127-158.
Kakes J. (1998), “Monetary transmission and business cycle asymmetry”, mimeo,
University of Groningen, September 1998.
Kashyap A, O. Lamont and J. Stein (1994), “Credit conditions and the cyclical
behavior of inventories”, Quarterly Journal of Economics, 59:3, p 565-592.
Kieler M. and T. Saarenheimo (1998), “Differences in monetary policy
transmission? A case not closed”, Economic Papers 132, Directorate General for
Economic and Financial Affairs.
Oliner S. and G. Rudebusch (1996), “Is there a broad credit channel for monetary
policy?”, Federal Reserve Bank of San Francisco Economic Review, No. 1, p 3-13.
Peersman G. and F. Smets (2001a), “The monetary transmission mechanism in the
euro area: more evidence from VAR analysis”, ECB Working Paper 91.
Peersman G. and F. Smets (2001b), “Are the effects of monetary policy in the euro
area greater in recessions than in booms?”, ECB Working Paper 52, forthcoming in
Mahadeva, L. and P. Sinclair (eds.), Monetary transmission in diverse economies,
Cambridge University Press, Cambridge, UK.
Vermeulen P. (2002), “Business fixed investment: evidence of a financial accelerator in
Europe”. Oxford Bulletin of Economics and Statistics, 64:3, July.
ECB • Working Paper No 165 • August 2002
33
(XURSHDQ&HQWUDO%DQN:RUNLQJ3DSHU6HULHV
)RUDFRPSOHWHOLVWRI:RUNLQJ3DSHUVSXEOLVKHGE\WKH(&%SOHDVHYLVLWWKH(&%·VZHEVLWH
KWWSZZZHFELQW
´)LQDQFLDOIULFWLRQVDQGWKHPRQHWDU\WUDQVPLVVLRQPHFKDQLVPWKHRU\HYLGHQFHDQGSROLF\
LPSOLFDWLRQVµE\&%HDQ-/DUVHQDQG.1LNRORY-DQXDU\
´0RQHWDU\WUDQVPLVVLRQLQWKHHXURDUHDZKHUHGRZHVWDQG"µE\,$QJHORQL$.DVK\DS
%0RMRQ'7HUOL]]HVH-DQXDU\
´0RQHWDU\SROLF\UXOHVPDFURHFRQRPLFVWDELOLW\DQGLQIODWLRQDYLHZIURPWKHWUHQFKHVµ
E\$2USKDQLGHV'HFHPEHU
´5HQWLQGLFHVIRUKRXVLQJLQ:HVW*HUPDQ\WRµE\-+RIIPDQQDQG&.XU]
-DQXDU\
´+HGRQLFKRXVHSULFHVZLWKRXWFKDUDFWHULVWLFVWKHFDVHRIQHZPXOWLXQLWKRXVLQJµE\
2%RYHUDQG39HOLOOD-DQXDU\
´'XUDEOHJRRGVSULFHLQGH[HVDQGTXDOLW\FKDQJHDQDSSOLFDWLRQWRDXWRPRELOHSULFHVLQ
,WDO\µE\*07RPDW-DQXDU\
´0RQHWDU\SROLF\DQGWKHVWRFNPDUNHWLQWKHHXURDUHDµE\1&DVVRODDQG&0RUDQD
-DQXDU\
´/HDUQLQJVWDELOLW\LQHFRQRPLFVZLWKKHWHURJHQRXVDJHQWVµE\6+RQNDSRKMDDQG.0LWUD
-DQXDU\
´1DWXUDOUDWHGRXEWVµE\$%H\HUDQG5($)DUPHU)HEUXDU\
´1HZWHFKQRORJLHVDQGSURGXFWLYLW\JURZWKLQWKHHXURDUHDµE\)9LMVHODDUDQG5$OEHUV
)HEUXDU\
´$QDO\VLQJDQGFRPELQLQJPXOWLSOHFUHGLWDVVHVVPHQWVRIILQDQFLDOLQVWLWXWLRQVµE\(7DEDNLV
DQG$9LQFL)HEUXDU\
´0RQHWDU\SROLF\H[SHFWDWLRQVDQGFRPPLWPHQWµE\*:(YDQVDQG6+RQNDSRKMD
)HEUXDU\
´'XUDWLRQYROXPHDQGYRODWLOLW\LPSDFWRIWUDGHVµE\60DQJDQHOOL)HEUXDU\
´2SWLPDOFRQWUDFWVLQDG\QDPLFFRVWO\VWDWHYHULILFDWLRQPRGHOµE\&0RQQHWDQG
(4XLQWLQ)HEUXDU\
´3HUIRUPDQFHRIPRQHWDU\SROLF\ZLWKLQWHUQDOFHQWUDOEDQNIRUHFDVWLQJµE\6+RQNDSRKMD
DQG.0LWUD)HEUXDU\
´2SHQQHVVLPSHUIHFWH[FKDQJHUDWHSDVVWKURXJKDQGPRQHWDU\SROLF\µE\)6PHWVDQG
5:RXWHUV)HEUXDU\
34
ECB • Working Paper No 165 • August 2002
´1RQVWDQGDUGFHQWUDOEDQNORVVIXQFWLRQVVNHZHGULVNVDQGFHUWDLQW\HTXLYDOHQFHµE\
$DO1RZDLKLDQG/6WUDFFD0DUFK
´+DUPRQL]HGLQGH[HVRIFRQVXPHUSULFHVWKHLUFRQFHSWXDOIRXQGDWLRQVµE\('LHZHUW
0DUFK
´0HDVXUHPHQWELDVLQWKH+,&3ZKDWGRZHNQRZDQGZKDWGRZHQHHGWRNQRZ"µE\
0$:\QQHDQG'5RGUtJXH]3DOHQ]XHOD0DUFK
´,QIODWLRQG\QDPLFVDQGGXDOLQIODWLRQLQDFFHVVLRQFRXQWULHVD´QHZ.H\QHVLDQµ
SHUVSHFWLYHµE\2$UUDWLEHO'5RGUtJXH]3DOHQ]XHODDQG&7KLPDQQ0DUFK
´&DQFRQILGHQFHLQGLFDWRUVEHXVHIXOWRSUHGLFWVKRUWWHUPUHDO*'3JURZWK"µE\
$0RXURXJDQHDQG05RPD0DUFK
´7KHFRVWRISULYDWHWUDQVSRUWDWLRQLQWKH1HWKHUODQGVµE\%%RGHDQG
-9DQ'DOHQ0DUFK
´7KHRSWLPDOPL[RIWD[HVRQPRQH\FRQVXPSWLRQDQGLQFRPHµE\)'H)LRUHDQG
37HOHV$SULO
´5HWDLOEDQNLQWHUHVWUDWHSDVVWKURXJKWKHQHZHYLGHQFHDWWKHHXURDUHDOHYHOµE\
*GH%RQGW$SULO
´(TXLOLEULXPELGGLQJLQWKHHXURV\VWHP·VRSHQPDUNHWRSHUDWLRQVµE\8%LQGVHLO$SULO
´1HZµYLHZVRQWKHRSWLPXPFXUUHQF\DUHDWKHRU\ZKDWLV(08WHOOLQJXV"µE\
)30RQJHOOL$SULO
´2QFXUUHQF\FULVHVDQGFRQWDJLRQµE\0)UDW]VFKHU$SULO
´3ULFHVHWWLQJDQGWKHVWHDG\VWDWHHIIHFWVRILQIODWLRQµE\0&DVDUHV0D\
´$VVHWSULFHVDQGILVFDOEDODQFHVµE\)(VFKHQEDFKDQG/6FKXNQHFKW0D\
´0RGHOOLQJWKHGDLO\EDQNQRWHVLQFLUFXODWLRQLQWKHFRQWH[WRIWKHOLTXLGLW\PDQDJHPHQWRI
WKH(XURSHDQ&HQWUDO%DQNµE\$&DEUHUR*&DPED0HQGH]$+LUVFKDQG)1LHWR
0D\
´$QRQSDUDPHWULFPHWKRGIRUYDOXLQJQHZJRRGVµE\,&UDZIRUG0D\
´$IDLOXUHLQWKHPHDVXUHPHQWRILQIODWLRQUHVXOWVIURPDKHGRQLFDQGPDWFKHGH[SHULPHQW
XVLQJVFDQQHUGDWDµE\06LOYHUDQG6+HUDYL0D\
´7RZDUGVDQHZHDUO\ZDUQLQJV\VWHPRIILQDQFLDOFULVHVµE\0)UDW]VFKHUDQG0%XVVLHUH
0D\
´&RPSHWLWLRQDQGVWDELOLW\²ZKDW·VVSHFLDODERXWEDQNLQJ"µE\(&DUOHWWLDQG3+DUWPDQQ
0D\
ECB • Working Paper No 165 • August 2002
35
´7LPHWREXLOGDSSURDFKLQDVWLFN\SULFHVWULFN\ZDJHRSWLPL]LQJPRQHWDU\PRGHOE\
0&DVDUHV0D\
´7KHIXQFWLRQDOIRUPRI\LHOGFXUYHVµE\9%URXVVHDX0D\
´7KH6SDQLVKEORFNRIWKH(6&%PXOWLFRXQWU\PRGHOµE\$(VWUDGDDQG$:LOOPDQ0D\
´(TXLW\DQGERQGPDUNHWVLJQDOVDVOHDGLQJLQGLFDWRUVRIEDQNIUDJLOLW\µE\5*URSS
-9HVDODDQG*9XOSHV-XQH
´*LQIODWLRQIRUHFDVWVµE\)&DQRYD-XQH
´6KRUWWHUPPRQLWRULQJRIILVFDOSROLF\GLVFLSOLQHµE\*&DPED0HQGH]DQG$/DPR-XQH
´(XURDUHDSURGXFWLRQIXQFWLRQDQGSRWHQWLDORXWSXWDVXSSO\VLGHV\VWHPDSSURDFKµE\
$:LOOPDQ-XQH
´7KHHXUREORFWKHGROODUEORFDQGWKH\HQEORFKRZPXFKPRQHWDU\SROLF\LQGHSHQGHQFH
FDQH[FKDQJHUDWHIOH[LELOLW\EX\LQDQLQWHUGHSHQGHQWZRUOG"µE\0)UDW]VFKHU-XQH
´<RXWKXQHPSOR\PHQWLQWKH2(&'GHPRJUDSKLFVKLIWVODERXUPDUNHWLQVWLWXWLRQVDQG
PDFURHFRQRPLFVKRFNVµE\-)-LPHQRDQG'5RGULJXH]3DOHQ]XHOD-XQH
´,GHQWLI\LQJHQGRJHQRXVILVFDOSROLF\UXOHVIRUPDFURHFRQRPLFPRGHOVµE\--3HUH]
DQG3+LHEHUW-XO\
´%LGGLQJDQGSHUIRUPDQFHLQUHSRDXFWLRQVHYLGHQFHIURP(&%RSHQPDUNHWRSHUDWLRQVµ
E\.*1\ERUJ8%LQGVHLODQG,$6WUHEXODHY-XO\
´4XDQWLI\LQJ(PERGLHG7HFKQRORJLFDO&KDQJHµE\36DNHOODULVDQG'-:LOVRQ-XO\
´2SWLPDOSXEOLFPRQH\µE\&0RQQHW-XO\
´0RGHOXQFHUWDLQW\DQGWKHHTXLOLEULXPYDOXHRIWKHUHDOHIIHFWLYHHXURH[FKDQJHUDWHµE\
&'HWNHQ$'LHSSH-+HQU\&0DULQDQG)6PHWV-XO\
´7KHRSWLPDODOORFDWLRQRIULVNVXQGHUSURVSHFWWKHRU\µE\/6WUDFFD-XO\
´3XEOLFGHEWDV\PPHWULHVWKHHIIHFWRQWD[HVDQGVSHQGLQJLQWKH(XURSHDQ8QLRQµE\
6.URJVWUXS$XJXVW
´7KHUDWLRQDOLW\RIFRQVXPHUV·LQIODWLRQH[SHFWDWLRQVVXUYH\EDVHGHYLGHQFHIRUWKHHXUR
DUHDµE\0)RUVHOOVDQG*.HQQ\$XJXVW
´(XURDUHDFRUSRUDWHGHEWVHFXULWLHVPDUNHWILUVWHPSLULFDOHYLGHQFHµE\*GH%RQGW
$XJXVW
36
ECB • Working Paper No 165 • August 2002
´7KHLQGXVWU\HIIHFWVRIPRQHWDU\SROLF\LQWKHHXURDUHDµE\*3HHUVPDQDQG)6PHWV
$XJXVW
ECB • Working Paper No 165 • August 2002
37